sakuralang
AGENTS.md4 starsChanged 24 days agoApache-2.0View on GitHub
Sakura Lang is an agent-first language. v1.0.0 is its first stable release (STABILITY.md); the native backend, packages and std/worker are experimental. Read CONTRIBUTING.md, docs/design.md, and STABILITY.md. Research is evidence; design is provisional. A future docs/spec/ will be normative. CODEOWNERS names reviewers; ADR 0002 maps the monorepo. The latest status and next steps are in the 2026-10-19 handoff brief. - Read directory guidance, especially website/AGENTS.md. Inspect helpers before adding code; preserve others' work and stay in scope. Avoid speculative features, abstractions,…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository7 other files it gives its agents.
- compiler/AGENTS.mdAGENTS.md
- skills/sakura/assets/AGENTS.mdAGENTS.md
- website/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skills/sakura/assets/CLAUDE.mdCLAUDE.md
- docs/llms.txtllms.txt
- sakuraSkill
The file
We split this AGENTS.md into the rules it gives an agent
5 rules. About 1 is common, 2 are distinctive, and 2 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 5
Read CONTRIBUTING.md, docs/design.md, and STABILITY.md.
- Distinctive: line 12
Read directory guidance, especially website/AGENTS.md.
- Distinctive: line 14
Avoid speculative features, abstractions, and dependencies.
- Very specific: line 34
Use RFCs for language/protocol/stability changes and ADRs for architecture.
- Common: line 40
Never expose credentials. Do not publish packages, deploy, send messages, or change remote settings without task authorization.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Working on Sakura Lang
Sakura Lang is an agent-first language. v1.0.0 is its first stable release
(STABILITY.md); the native backend, packages and std/worker are experimental.
Read CONTRIBUTING.md, docs/design.md, and
STABILITY.md. Research is evidence; design is provisional. A future
docs/spec/ will be normative. CODEOWNERS names reviewers;
ADR 0002 maps the monorepo.
The latest status and next steps are in
the 2026-10-19 handoff brief.
- Read directory guidance, especially website/AGENTS.md. Inspect helpers before adding code; preserve others' work and stay in scope. Avoid speculative features, abstractions, and dependencies.
- Website: Bun 1.4.3 (manifest pin). In
website/, runbun install --frozen-lockfile, thenbun run fmt:check,bun run lint(oxlint),bun run typecheckandbun run build;bun run fmtformats (oxfmt) andbun run devserves locally. - Compiler: the Cargo workspace lives in
compiler/, with its development toolchain pinned incompiler/rust-toolchain.toml. Fromcompiler/, runcargo fmt --all -- --check,cargo clippy --workspace --all-targets --locked -- -D warnings, andcargo test --workspace --locked. These are current gates; report their results. The required MSRV CI job readsworkspace.package.rust-versionfrom the manifest and checks all workspace targets and runs workspace tests on that toolchain. - Map checks to changes: docs → links/examples; website → lint/build and affected UI; workflows → YAML/actionlint; compiler → targeted regression and conformance cases plus workspace checks. Require deterministic output, malformed-input recovery, explicit resource limits, and versioned diagnostic contracts.
- Generated website
.source/and.next/come from Fumadocs/Next.js tooling; edit their source files. Preserve lockfiles. The compiler has JS and WASM code generation, fixture harnesses and checked-in snapshots. Review every snapshot diff; never normalize away semantics. Add spec-linked positive and negative cases. - Use RFCs for language/protocol/stability changes and ADRs for architecture. Explain compatibility and update the changelog for user-visible changes. Label pseudocode and unimplemented goals.
- Disclose AI involvement; the human contributor owns the result. Report exact checks and limitations, never fabricated passes. Obtain named maintainer review for semantics/security changes; do not weaken contracts to make checks pass.
- Never expose credentials. Do not publish packages, deploy, send messages, or change remote settings without task authorization. Explicitly report checks that could not run. Use Conventional Commit PR titles and focused commits.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

